Desensitized CMOS Low-Noise Amplifiers

Abstract

The minimum attainable noise figure for scaled- CMOS low-noise amplifiers (LNAs) is limited by impedance mismatches such as the well-known noise/power tradeoff. In this paper, we show that a power-constrained optimization of the device noise resistance parameter, R <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">n</sub> , significantly reduces the impact of mismatches and variations and leads to an almost simultaneous noise and power match. This process, called desensitization, makes the design largely immune to measurement and modeling errors and manufacturing variations, and significantly reduces frequency-dependent noise mismatches in wide-band LNAs. Measured data from devices and desensitized LNAs designed on 180-nm and 90-nm CMOS processes shows that: (1) a device size selected for optimum R <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">n</sub> middot is less sensitive to source impedance mismatches and provides a wide-band noise match; and (2) LNAs approach a simultaneous input and noise match, and exhibit significant improvements (ges 2x) in their wide-band noise performance.